Andrew
Trotman, Tim Jones, and Chris Handley
Players
downloading GPS coordinates from the internet, hiking to the given spot, and
hunting for a hidden box – this is the new sport of geocaching. Today there are
nearly 200,000 such boxes in over 200 countries. With so many to find, a
recommender is needed, one that takes into account not only the boxes, but also
the geospatial and temporal nature of the sport.
A database of
geocaches in the South Island of New Zealand is made by trawling a prominent
geocaching web site. This is then used to estimate the home-coordinates (geospatial
playing centre) of players. Predictions are verified against a set of correct
coordinates solicited from players.
Several geocache
recommenders are discussed and compared. The precision, computed using mean of
mean reciprocal rank (MMRR), of each is measured. The best method tried is a
collaborative filter using intersection over mean to find similar players and a
voting scheme to recommend geocaches. This method is proposed as a replacement
for the currently used distance from home-coordinate; doing so will increase
the precision of existing systems such as geocaching.com.